Cryptographic primitives usually assert some security level given as number of operations to mount an attack. Hash functions, for example, give different security levels for collision attacks, preimage attacks and second preimage attacks. From these, "safe" key sizes are derived for different primitives.
There are many different recommendations for safe key sizes and many different means of estimating future capabilities in performing computation. For example, www.keylength.com has a lot of these recommendations combined.
What I'm looking for, however, is the amount of simple operations that can be obviously seen as out of reach for all humanity for the foreseeable future - or actually, the lowest such value that is still believable.
It is very obvious that 2^256 simple operations is something that will never be reached. It is also very obvious that 2^64 simple operations can be reached as it already has been. Many of the recommendations seem to calculate 2^128 as a number that would be safe for 30 years or more. So the value I am looking for is likely between 2^128 and 2^256. I am guessing 2^160 or 2^192 might be safely out of reach.
But I want concrete arguments that can be easily reasoned about. I'd love to see arguments that are based on simple laws of physics or relations to concrete constants about the universe. For example, Landauer's principle could be used.
Note: the actual simple operations used are not relevant here - they might be operations on a quantum computer, or hash invocations, or whatever.